Data methodology
Where our numbers come from
We would rather show a checkable number than a confident-sounding guess, and we would rather say "we cannot tell" than invent a figure. Here is exactly how the HVAC Bench market pages are built.
The four figures
- Ad click cost is the average Google Ads cost-per-click for one named search term in one metro: what advertisers pay Google for a click from someone searching that term. It is the cost of a click, not the cost of a booked customer. Nobody knows your close rate but you, which is why the break-even calculator asks for it instead of assuming it.
- Search demand is how many times per month people search that term in that metro. It is search interest, not jobs sold.
- Technician wages are the median hourly wage for the trade's main occupation, with a low and high figure. US metros: the metro median and 10th and 90th percentile from the US Department of Labor. Canadian metros: Job Bank's median, low and high wage for the occupation's NOC unit group in the metro's economic region, in Canadian dollars; Job Bank's low and high are in most cases the 10th and 90th percentile. Where Job Bank shows a provincial figure for a region (Toronto), the page says so.
- Input costs are monthly price indexes for copper, steel, heating and cooling equipment, aluminum and diesel. US metros show US Bureau of Labor Statistics producer price indexes; Canadian metros show Statistics Canada industrial product price indexes. Both are national series and are labelled with their source and series id. A US series is never shown as a Canadian figure, or the other way round.
There is no job price range on this site. An earlier version showed one, and it was a hardcoded number with no source, presented on every metro page as if it were local. It was removed on 2026-09-14 and will not return unless it arrives from a named, reproducible source with a retrieval date.
Sources
Ad click cost and search demand come from DataForSEO, which resells Google Ads keyword data through its Keywords Data API (the google_ads/search_volume endpoint). Wages come from the US Department of Labor via CareerOneStop, which republishes Bureau of Labor Statistics Occupational Employment and Wage Statistics by metropolitan area. Input costs come from the Bureau of Labor Statistics Producer Price Index API. In Canada, wages come from Job Bank open wage data (Employment and Social Development Canada) and input costs from Statistics Canada table 18-10-0266-01, Industrial product price index by product. Every published row carries its source and the date it was retrieved, and the CSV download includes both.
One search term per job
Each job on the market page is measured by exactly one search term, and that term is printed under the job name. Ad click cost, search demand and the monthly history for a job all come from the same row of the same vendor response, so any figure we publish can be explained in one sentence: "this is what 'ac repair' did in Phoenix". We request related phrasings as well and keep them for audit, but nothing from them reaches a published number. An earlier version blended several phrasings into one cost figure, which meant a dead phrase with a junk price could drag a real one. That is gone.
What we hold back
Every refresh lands as a candidate and is checked before anything reaches a page. Impossible values (a negative volume, a month that does not exist) block the refresh outright. Suspicious values that could still be real, such as a click cost under fifty cents on a paid trade term, or a term the vendor returned no cost for, are quarantined: kept, logged with the reason, and not shown. After that, a metro click cost under 15% or over four times its country's national figure for the same term is held from display the same way. Click costs on terms with fewer than 100 searches a month are shown but marked low volume, and they do not count toward a page's headline comparison. A missing cell on a market page means we had a number and did not trust it, not that we forgot to fill it in. Where Google returns no click cost for a term in a metro, the cell may show the click cost for the same term across the whole state or province, from the same source, labelled with its region. It is never written into the metro cell and never used by the calculator.
Seasonality
Where we have monthly history we show an indexed profile: each month as a percentage of that term's annual average, so 200 means twice the average month. We label how much to trust it. Strong means the shape repeats across two or more years. Directional means we have one year only, and a single year's weather can look like a pattern. Where history is too short or too flat we say we cannot tell. We do not publish a peak-to-trough ratio: at low volumes Google rounds monthly figures into coarse buckets and a ratio of two rounded buckets is not a finding. A quiet month that reads zero is a legitimate answer from a low-volume market and is shown as such.
Metros and geography
Search figures are pulled per metropolitan area using Google Ads location targeting, which follows Google's own metro boundaries. Wages follow the Bureau of Labor Statistics metropolitan statistical area. The two do not line up exactly, and in large markets such as New York or Chicago the advertising footprint can be materially larger than the statistical one. Each is labelled with what it covers. Metro-level search volumes are naturally much smaller than national ones: a few hundred monthly searches for a specific service in one city is normal, not an error.
Canadian metros show ad click cost and search demand in US dollars, because that is the currency the vendor reports in, and every figure says so. Wages are in Canadian dollars, as Job Bank publishes them, and input costs are Statistics Canada indexes, which carry no currency.
Freshness
Search figures (click cost and demand) are pulled once a year, and each pull carries up to four years of monthly history, which is what the seasonal profiles are built from. Every page shows the retrieval date. Input costs update monthly with the BLS and Statistics Canada releases. Wages are published annually (Job Bank's latest edition is November 2025).
What it is not
These are market signals, not a promise. Search demand is not a count of jobs available, cost-per-click is not what you will pay per booked customer, and nothing here is a quote. Use the numbers to sanity-check your own pricing and ad spend, not as the last word. Spot something that looks off? Tell us and we will check it.